SEARCH

SEARCH BY CITATION

Keywords:

  • Diabetes mellitus;
  • glucose curve;
  • hyperglycemia;
  • hypoglycemia

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. References

Background

Portable blood glucose meters (PBGMs) are useful for serial measurements of blood glucose and creation of blood glucose curves in veterinary practice. However, it is necessary to validate PBGMs designed for people for veterinary use.

Objectives

Our objective was to evaluate the accuracy of 2 PBGMs designed for people for use in dogs and cats.

Methods

The blood glucose levels were determined in blood samples collected from 69 dogs and 26 cats admitted to the Kagoshima University Veterinary Teaching Hospital, using a MEDISAFE [PBGM-T] and an Antsense III [PBGM-H], and a FUJI DRI-CHEM 7000V as reference method. The correlations and agreements among the results were statistically analyzed.

Results

Simple regression analyses revealed a high correlation between values from both PBGMs and the reference method in both dogs and cats. However, Passing–Bablok regression and Bland–Altman analyses revealed that the data from both PBGMs did not show statistical agreement with the reference values. Concordance correlated coefficients were moderate for the PBGM-T and almost perfect for the PBGM-H for canine samples, and were poor for the PBGM-T and substantial for the PBGM-H for feline samples. Hematocrit values significantly affected the results of the PBGM-T, but not the PBGM-H. Error grid analyses revealed that all measurements from both PBGMs would lead to acceptable treatment decisions.

Conclusions

Our findings suggest that both PBGMs, especially the PBGM-H, would be clinically useful in small animal practice, although there was a bias between each PBGM and the reference method.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. References

In canine and feline veterinary practice, blood glucose measurements are often required to make diagnoses and monitor patients. The creation and evaluation of a blood glucose curve are essential for the diagnosis and management of diabetes mellitus.[1] However, frequent blood sampling is required and, if an automated chemistry analyzer is used, a relatively large volume of blood is needed, necessitating the use of a needle and syringe for blood collection. This results in both physical and emotional stress for the animals. In addition, although measurement of blood glucose using an automated chemistry analyzer provides accurate results, the preparation of plasma or serum samples and the analysis itself are time consuming. To reduce animal stress and the complexity of measurement procedures, a variety of portable blood glucose meters (PBGMs) have been recently assessed for the easy creation of blood glucose curves for dogs and cats.[2-11]

An advantage of most PBGMs is that they require only a small blood sample to quantitate blood glucose concentrations. Another major advantage is that the results are obtained within a few seconds of sample collection.[2, 5, 7, 8] This is important for creating a blood glucose curve and for the management of critical conditions. Furthermore, most PBGMs are easy to handle. Therefore, they are well-suited for home monitoring of blood glucose by the owners of dogs or cats with diabetes mellitus.[10, 11] Despite these advantages, the accuracy of PBGM in comparison with an automated analyzer is sometimes questioned;[2, 3] therefore, careful validation is necessary before using PBGMs in canine and feline veterinary practice.

In this study, we compared the blood glucose levels of dogs and cats using 2 types of PBGMs and an automated chemistry analyzer. The PBGMs that we tested are currently used by some veterinarians in clinical practice, but have not been previously evaluated for animal use.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. References

Blood samples

A total of 69 client-owned dogs and 26 cats were admitted to the Kagoshima University Veterinary Teaching Hospital (KUVTH) for both disease evaluation and wellness checks. From these animals, 111 canine and 40 feline blood samples were collected for this study between September and October 2011. The blood samples were collected from the jugular or cephalic vein and were anti-coagulated with EDTA or heparin. A CBC was performed using the EDTA-treated blood samples. Blood glucose was assessed in the heparinized whole-blood samples using 2 PBGMs (see details below). Remaining heparinized samples were centrifuged and plasma was collected within 5 minutes of blood collection. Blood glucose levels in the plasma samples were quantified using an automated chemistry analyzer (details see below). Samples were included whether patients were hypo-, eu- or hyperglycemic, but cases with data missing for the CBC or any of the 3 blood glucose measurements were excluded from this prospective study. In addition, samples showing severe hemolysis, lipemia, or icterus were also excluded to avoid confounding factors.

The automated chemistry analyzer and PBGMs

In general, blood glucose values are measured in analyzers that use the hexokinase method to determine glucose levels. In this study, an automated chemistry analyzer that uses colorimetric detection of the glucose oxidase (GOD)/peroxidase (POD) reaction (FUJI DRI-CHEM 7000V, FUJIFILM, Tokyo, Japan) was used to determine the reference blood glucose level. The FUJI DRI-CHEM is routinely used in KUVTH, and in-house data from the manufacturer show that it yields results that are almost identical to those obtained using the hexokinase method. The correlation coefficient of blood glucose levels measured by the hexokinase method and FUJI DRI-CHEM was 0.999, and coefficients of variation (CVs) for glucose in dogs and cats were 1.1% and 1.4%, respectively. The required sample volume is 10 μL of plasma and the linear range is 10 to 600 mg/dL.

Two PBGMs were used in this study: the MEDISAFE Mini (TERUMO Corporation, Tokyo, Japan) (PBGM-T) and the Antsense III (Horiba Medical, Kyoto, Japan) (PBGM-H). The PBGM-T is designed for the self-monitoring of blood glucose by human diabetes patients and measures blood glucose levels in whole blood via colorimetric detection of the GOD/POD reaction. Minimum sample volume is 1.2 μL of whole blood and the linear range is 20 to 600 mg/dL. It takes 10 s to obtain a result. The PBGM-H is usually used for point-of-care testing and uses the enzyme electrode method to detect hydrogen peroxide produced by the GOD reaction. A 5 μL whole-blood sample is applied directly to a semi-permeable membrane of the cartridge, which retains blood cells but allows plasma to diffuse through the membrane containing immobilized GOD. Hydrogen peroxide produced by the GOD reaction stimulates the electrical current at the electrode, the differential value of which is used to calculate the blood glucose level. The linear range is 10–1000 mg/dL and the result is obtained 45 seconds after sample application.

Statistical analyses

The blood glucose value obtained from the automated chemistry analyzer was considered the reference standard glucose concentration for each sample. The values from both PBGMs were compared with the reference standard values. Method comparison was carried out as described by Jensen and Kjelgaard-Hansen.[12] Statistical analyses were conducted using Medcalc statistical software version 12.2.1 (MedCalc Software, Mariakerke, Belgium).

Correlation coefficients (r) between the values obtained from the reference method and from each PBGM were calculated, and were interpreted as follows: 0.90–1.00 = very high correlation; 0.70–0.89 = high correlation; 0.50–0.69 = moderate correlation; 0.30–0.49 = low correlation; and 0–0.29 = little correlation.

Passing–Bablok linear regression analysis was used to determine constant and/or proportional differences between the analytical methods as has been previously described.[13] Constant differences were detected by calculation of the intercept of the regression equation within the 95% confidence interval (CI). If the CI for the intercept did not contain the value 0, then results obtained using the PBGM were considered to differ from those obtained by the reference method by a constant amount. In addition, if the slope of the regression equation did not contain the value 1 within its 95% CI, it was considered to indicate a proportional difference between the 2 methods of analysis.

Bland–Altman difference plot analyses were established to compare each of the 2 instruments, PBGM-T or PBGM-H, with the reference method. This method plots the difference between the 2 results against the respective averages of all results. CVs for glucose were obtained from the manufacturer for the reference method (dog: 1.1%; cat: 1.4%) as described above. CVs for PBGM-T and -H were calculated by repetitive analyses using blood samples from 2 dogs and 2 cats. Blood samples were anti-coagulated with sodium fluoride and glucose concentrations of each sample were measured 10 times using the PBGM-T or PBGM-H. Glucose concentrations in these samples were determined with the FUJI DRI-CHEM and ranged from 90 to 100 mg/mL. All measurements were performed at room temperature (22°C) and completed within a 15-minute time period. From these analyses, CVs were determined for PBGM-T (dog: 2.8% and cat: 2.5%) and PBGM-H (dog: 2.7% and cat: 3.0%). The combined inherent imprecision of the reference method and the PBGM was calculated as CVBoth Methods =(CV2Ref + CV2PBGM)0.5.[12] In addition, the acceptance limit was calculated using the following formula: Acceptance limit = 0 ± (1.96 × CVBoth Methods ×AverageBoth Methods).

Concordance correlation coefficients were calculated to estimate the strength of the relationship between the reference method and each PBGM.[14] Concordance correlation coefficient values (pc) were interpreted as follows: > 0.99: almost perfect concordance; 0.95–0.99: substantial concordance; 0.90–0.95: moderate concordance; and < 0.90: poor concordance.

The effect of the HCT value on the PBGM results was analyzed by scatter plot analysis. The absolute value of a difference (y-axis) in measured blood glucose between the reference method and each PBGM was plotted against its own HCT (x-axis). The influence of HCT values on measured blood glucose levels from each PBGM was statistically analyzed by calculating the correlation coefficients between HCT values and the absolute differences between the reference method and each PBGM. Correlation coefficients were assessed as described above.

Error grid analyses

The clinical relevance of the data from each PBGM was evaluated using error grid analysis with the assumptions in use for diabetic dogs and cats as described elsewhere.[2, 4, 8, 9, 15] Briefly, the 5 grids (zones A to E) were assigned based on the PBGM glucose values (y-axis) vs the glucose values obtained using the reference method (x-axis). PBGM results that deviated from the reference value by less than 20% were plotted in zone A. If both the PBGM and the reference method results were < 70 mg/dL, the results were also plotted in zone A. The data in zone A are considered to be clinically accurate and likely to lead to clinically correct treatment decisions. Zone B is defined as containing data with errors of more than 20%, but without influence on the clinical treatment decision. Values plotted in zones C, D and E are those that may lead to treatment errors or failure.[2, 4] Specifically, values in the upper zone C would lead to unnecessary correction of acceptable blood glucose concentrations and might lead to iatrogenic hypoglycemia if insulin was administered to correct the apparent hyperglycemia, whereas values in lower zone C might lead to hyperglycemia because of the resulting reduced insulin doses. Zone D represents potentially dangerous errors, where the PBGM yielded glucose values within the clinical target range, while the correct values are outside this range. The PBGM values in zone E are opposite to the correct glucose values. In zone E, correct hyperglycemia and hypoglycemia determined by the reference method correspond to hypoglycemia and hyperglycemia determined by the PBGM, respectively. Readings in this zone may lead to inappropriate and life-threatening changes in insulin treatment.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. References

No samples were outside of the measurement range (ie, expressed on the PBGMs as “high” or “low”), so numeric blood glucose concentrations were obtained for all samples. Therefore, all data from the analyzed samples were included in this study. The blood glucose concentrations in 111 canine blood samples measured with the PBGM-T and PBGM-H ranged from 22 to 455 mg/dL and from 50 to 415 mg/dL, respectively. The reference method yielded similar results, with blood glucose concentrations ranging from 54 to 395 mg/dL. In 40 feline blood samples, blood glucose concentrations ranged from 30 to 401 mg/dL and from 56 to 410 mg/dL when tested with the PBGM-T and PBGM-H, respectively. Corresponding values from the reference method ranged from 58 to 392 mg/dL. There was a high correlation between the values of the PBGM-T and the reference method (r = .820, < .001 and r = .844, < .001, for canine and feline samples, respectively). There was also a very high correlation between values obtained using the PBGM-H and the reference values (r = .941, < .001 and r = .934, < .001 for canine and feline samples, respectively). However, an additional regression analysis was required to estimate constant and proportional errors, because all correlation coefficients calculated by ordinary least squares regression were less than 0.975.

In dogs, Passing–Bablok regression analysis of PBGM-T values against reference values yielded an intercept of −51.6 (95% CI, −62.8 to −42.0) and slope of 1.28 (95% CI, 1.20–1.36, Figure 1A), indicating a proportional and a constant bias. Likewise, there were a proportional and a constant bias, with an intercept of −6.5 (95% CI, −10.7 to −1.9) and a slope of 1.06 (95% CI, 1.02–1.10, Figure 1B) when PBGM-H values were compared with the reference method. In cats, Passing-Bablok regression of PBGM-T values against reference values showed an intercept of −47.5 (95% CI, −65.4 to −32.4) and a slope of 1.14 (95% CI, 1.01–1.26), suggesting both proportional and constant biases (Figure 1C). However, the regression of PBGM-H values against reference values yielded an intercept of −5.2 (95% CI, −12.3 to 3.8) and a slope of 1.10 (95% CI, 1.04–1.15), indicating a proportional inconstant bias (Figure 1D).

image

Figure 1. Scatter plots of blood glucose concentrations and Passing–Bablok regression analysis of blood glucose concentrations determined in canine and feline whole-blood samples with a MEDISAFE Mini (PBGM-T, y-axis), and an Antsense III (PBGM-H, y-axis) portable blood glucose meter, and a FUJI DRI-CHEM as reference analyzer (x-axis). The dotted line represents the theoretical line of equality with the reference method. Fitted lines and approximation formulas from Passing–Bablok regression analyses are presented in each graph. (A) and (C) show results in canine and feline samples measured by PBGM-T, and (B) and (D) show results in canine and feline samples measured by PBGM-H.

Download figure to PowerPoint

Bland–Altman plot analysis revealed a mean difference of −21.2 mg/dL between the PBGM-T and the reference method (Figure 2A) and a mean difference of 1.4 mg/dL between the PBGM-H and the reference method (Figure 2B) in canine samples. For the PBMG-T, most of the data (96/111 samples, 86.5%) were beyond the acceptable limit calculated from the combined inherent imprecision of the PBGM-T and the automated analyzer (Figure 2A). Conversely, 84.7% of the samples (94/111 samples) measured using the PBGM-H were plotted within the acceptable limits (Figure 2B). Likewise, feline blood samples revealed a mean difference of −34.7 mg/dL between the PBGM-T and the reference method (Figure 2C), and a mean difference of 8.8 mg/dl between the PBGM-H and reference method (Figure 2D). While only 3/40 feline samples (7.5%) measured with the PBGM-T had blood glucose concentrations that fell within acceptable limits, only 52.5% of the data (21/40 samples) from the PBGM-H were within acceptable limits.

image

Figure 2. Bland–Altman difference plots of differences in blood glucose concentrations determined in canine and feline whole-blood samples with a MEDISAFE Mini (PBGM-T) or an Antsense III (PBGM-H) portable blood glucose meter, and a FUJI DRI-CHEM as reference analyzer (y-axis), plotted against the respective average concentrations determined with each method (x-axis). Solid lines represent the theoretical line of equality with the reference method. Dashed lines represent 0 ± (1.96 × CVBoth Methods × AverageBoth Methods) and their intervals represent acceptable limits. The dotted line represents the mean difference between the 2 methods. (A) and (C) show results in canine and feline samples measured by PBGM-T, and (B) and (D) show results in canine and feline samples measured by PBGM-H.

Download figure to PowerPoint

The pc values for canine samples were 0.90 and 0.99 for PBGM-T and -H, respectively. Therefore, the concordance correlated coefficients were moderate for the PBGM-T and almost perfect for the PBGM-H. For feline samples, the pc values were 0.83 and 0.97 for PBGM-T and -H results, respectively. Thus, concordance correlated coefficients were poor for PBGM-T and substantial for PBGM-H.

There was a moderate correlation between canine HCT values and mean differences for the PBGM-T (r = .5662, < .0001), and little correlation between HCT values and mean differences for the PBGM-H (r = −.2767, = .0032, Figure 3A,B). In feline samples, the correlation was moderate between HCT values and mean differences for the PBGM-T (r = .6050, < .0001), while HCT and errors for the PBGM-H were not significantly correlated (r = −.2745, = .0866, Figure 3C,D).

image

Figure 3. Scatter plots of the absolute differences (y-axis) between blood glucose concentrations determined in canine and feline whole-blood samples with a MEDISAFE Mini (PBGM-T) and an Antsense III (PBGM-H) portable blood glucose meter, and a FUJI DRI-CHEM as reference analyzer, plotted against the HCT values (x-axis). Fitted lines are depicted in each graph. (A) and (C) show results in canine and feline samples measured by PBGM-T, and (B) and (D) show results in canine and feline samples measured by PBGM-H.

Download figure to PowerPoint

Error grid analyses of the blood glucose concentrations in canine samples revealed that all measurements from both PBGMs were within zones A or B (Figures 4A,B). Fifty-nine (53.2%) and 110 (99.1%) out of 111 samples were plotted in zone A for the PBGM-T and PBGM-H, respectively. For feline samples, one measurement was plotted in zone C from the PBGM-T, but all other measurements were located in zones A (14/40, 35.0%) and B (25/40, 62.5%, Figure 4C). However, all concentrations determined by the PBGM-H were located in zones A (37/40, 92.5%) and B (3/40, 7.5%, Figure 4D).

image

Figure 4. Error grid analyses of blood glucose concentrations determined in canine and feline whole-blood samples with a MEDISAFE Mini (PBGM-T) and an Antsense III (PBGM-H) portable blood glucose meter, and a FUJI DRI-CHEM as reference analyzer. (A) and (C) show results in canine and feline samples measured by PBGM-T, and (B) and (D) show results in canine and feline samples measured by PBGM-H.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. References

Our regression analyses indicated that there were biases between the results from both PBGMs and the reference method, although the results from the PBGM-H were more similar to the reference values for both canine and feline samples than those from the PBGM-T.

The most important question that we answered with this study is whether PBGM designed for human blood samples could be used for clinical decision making in dogs and cats. Although biases between data from both PBGMs and the reference method were observed, our results in the error grid analyses revealed that both tested PBGMs yielded results leading to clinically acceptable treatment decisions. All of the data except for one measurement from the PBGM-T were plotted in zone A or B, suggesting that the clinical outcome would not be affected by the results obtained using the PBGMs tested in this study. It was especially noteworthy that 99.1% and 92.5% of the results from the PBGM-H in dogs and cats, respectively, were plotted in zone A. These numbers are not only higher than those from the PBGM-T but are also higher than the results from previous similar studies.[2, 3, 5, 9] Therefore, both PBGMs are considered to be applicable for small animal practice.

It is important to consider the effect of HCT on results from the PBGM-T. The user manual for the PBGM-T states that results are not affected by HCT values ranging from 20 to 60% in human blood. Most of our samples were within this range, but the results from the PBGM-T were significantly affected by HCT values in canine and feline samples. We were unable to determine the reason why HCT values affected the results. Factors such as intracellular glucose levels, blood viscosity and filtration efficacy of blood cells in the device may have influenced the results, but further studies are required.[16] As the PBGM-H is not affected, it appears more suitable for veterinary use.

There are some limitations to this study. It has been previously reported that the glucose level in capillary blood is approximately 20% to 25% higher than the glucose level in venous blood.[17] In clinical practice, PBGMs would be used to evaluate glucose levels in capillary blood obtained from the foot pad or ear pinna; therefore, it may be necessary to confirm that similar accuracies are obtained when using capillary blood samples. However, we have not evaluated this because the objective of this study was solely to compare the results from the reference method with the results obtained using PBGMs.

The second limitation is the sample population. We used multiple samples from the same animals to ensure an adequate number of samples in this study. Most statistical tests assume that samples are independent observations. Additional calculations were performed following removal of nonindependent samples (data not shown). In these calculations, the number of samples with out-of-range values for both PBGMs increased in Bland–Altman analyses and the effect of HCT values on results from the PBGM-T was exaggerated. Furthermore, our nonindependent samples were obtained from cases with at least a one-week interval between samples, or from diabetic cases undergoing blood glucose curve creation. Although our results might have been biased by the sample population, we consider that the results obtained in this study are more accurate than those from only independent samples.

As a third limitation of this study, hypoglycemic samples were underrepresented in this study and were identified in only 7 dogs and one cat. The distribution of the hypoglycemic samples was uneven. Although prolonged storage of blood samples would have allowed the generation of artificial hypoglycemia, we chose to analyze all samples within 5 minutes of collection to avoid glycolysis of plasma glucose and undesirable hemolysis. Hypoglycemic samples could also have been prepared by diluting the blood samples with saline or phosphate buffer, but as both PBGMs are intended to be used with whole blood, and the HGB concentration could have affected results, we chose not to dilute the samples. However, for a complete assessment of the correlation between the results generated by PBGMs in comparison with the reference method, more hypoglycemic samples need to be evaluated.

A fourth limitation of this study was the application of an error grid analysis designed for human patients. Previous studies have shown that using a human system for error grid analysis gives satisfactory results with canine and feline samples.[2, 4, 8, 9, 15] However, different results with dog- or cat-specific error grid systems cannot be excluded.[3]

Disclosure: The authors have indicated that they have no affiliations or financial involvement with any organization or entity with a financial interest in, or in financial competition with, the subject matter or materials discussed in this article.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. References
  • 1
    Feldman EC, Nelson RW. Diabetes mellitus. In: Feldman EC, Nelson RW, ed. Canine and Feline Endocrinology and Reproduction. 2nd ed. Philadelphia, PA: WB Saunders; 1996:339391.
  • 2
    Wess G, Reusch C. Assessment of five portable blood glucose meters for use in cats. Am J Vet Res. 2000;61:15871592.
  • 3
    Johnson BM, Fry MM, Flatland B, Kirk CA. Comparison of a human portable blood glucose meter, veterinary portable blood glucose meter, and automated chemistry analyzer for measurement of blood glucose concentrations in dogs. J Am Vet Med Assoc. 2009;235:13091313.
  • 4
    Dobromylskyj MJ, Sparkes AH. Assessing portable blood glucose meters for clinical use in cats in the United Kingdom. Vet Rec. 2010;167:438442.
  • 5
    Cohen TA, Nelson RW, Kass PH, Christopher MC, Feldman EC. Evaluation of six portable blood glucose meters for measuring blood glucose concentration in dogs. J Am Vet Med Assoc. 2009;235:276280.
  • 6
    Wess G, Reusch C. Evaluation of five portable blood glucose meters for use in dogs. J Am Vet Med Assoc. 2000;216:203209.
  • 7
    Cohn LA, McCaw DL, Tate DJ, Johnson JC. Assessment of five portable blood glucose meters, a point-of-care analyzer, and color test strips for measuring blood glucose concentration in dogs. J Am Vet Med Assoc. 2000;216:198202.
  • 8
    Ghys T, Goedhuys W, Spincemaille K, Gorus F, Gerlo E. Plasma-equivalent glucose at the point-of-care: evaluation of Roche Accu-Chek Inform and Abbott Precision PCx glucose meters. Clin Chim Acta. 2007;386:6368.
  • 9
    Wess G, Reusch C. Capillary blood sampling from the ear of dogs and cats and use of portable meters to measure glucose concentration. J Small Anim Pract. 2000;41:6066.
  • 10
    Casella M, Wess G, Reusch CE. Measurement of capillary blood glucose concentrations by pet owners: a new tool in the management of diabetes mellitus. J Am Anim Hosp Assoc. 2002;38:239245.
  • 11
    Casella M, Wess G, Hässig M, Reusch CE. Home monitoring of blood glucose concentration by owners of diabetic dogs. J Small Anim Pract. 2003;44:298305.
  • 12
    Jensen AL, Kjelgaard-Hansen M. Method comparison in the clinical laboratory. Vet Clin Pathol. 2006;35:276286.
  • 13
    Passing H, Bablok W. A new biometrical procedure for testing the equality of measurements from two different analytical methods. Application of linear regression procedures for method comparison studies in clinical chemistry, Part I. J Clin Chem Clin Biochem. 1983;21:709720.
  • 14
    Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics. 1989;45:255268.
  • 15
    Clarke WL, Cox D, Gonder-Frederick LA, Carter W, Pohl SL. Evaluating clinical accuracy of systems for self-monitoring of blood glucose. Diabetes Care. 1987;5:622628.
  • 16
    Magnani M, Dachà M, Bossù M, Fornaini G. Comparative studies on red blood cell glucose phosphorylating activities of mammals. Comp Biochem Physiol B. 1978;60:323327.
  • 17
    Sacks DB. Carbohydrates. In: Burtis CA, Ashwood ER, Bruns DE, eds. Tietz Textbook of Clinical Chemistry and Molecular Diagnostics. 4th ed. St. Louis, MO: Elsevier Saunders; 2006:837901.